Dominant Defects in Drosophila Eye Pigmentation Resulting From a Euchromatin-Heterochromatin Fusion Gene

Corresponding author: Kent G. Golic, Department of Biology, 201 Biology Bldg., University of Utah, Salt Lake City, UT 84112., golic{at}bioscience.utah.edu (E-mail).

Communicating editor: S. HENIKOFF

	ABSTRACT

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

We have isolated a dominant mutation, pugilist^Dominant (pug^D), that causes variegated reductions in pteridine and ommochrome pigmentation of the Drosophila eye. The effect of pug^D on pteridine pigmentation is most dramatic: the only remaining pigment consists of a thin ring of pigment around the periphery of the eye with a few scattered spots in the center. The pug^D mutation disrupts a gene that encodes a Drosophila homolog of the trifunctional enzyme methylenetetrahydrofolate dehydrogenase (MTHFD; E.C.1.5.1.5, E.C.3.5.4.9, E.C.6.3.4.3). This enzyme produces a cofactor that is utilized in purine biosynthesis. Because pteridines are derived from GTP, the pigment defect may result from an impairment in the production of purines. The mutant allele consists of a portion of the MTHFD coding region fused to ~1 kb of highly repetitive DNA. Transcription and translation of both parts are required for the phenotype. The repetitive DNA consists of ~140 nearly perfect repeats of the sequence AGAGAGA, a significant component of centric heterochromatin. The unusual nature of the protein produced by this gene may be responsible for its dominance. The repetitive DNA may also account for the variegated aspect of the phenotype. It may promote occasional association of the pug^D locus with centric heterochromatin, accompanied by inactivation of pug^D, in a manner similar to the proposed mode of action for brown^Dominant.

THE discovery of the white-eyed mutant by Morgan marked the advent of Drosophila as a genetic model organism. Since then, dozens of eye pigment mutants have been isolated in Drosophila melanogaster (LINDSLEY and ZIMM 1992 ). The existence of so many easily recognized variants in eye color fostered the development of biochemical genetics (BEADLE and EPHRUSSI 1935 , BEADLE and EPHRUSSI 1936 ) and significantly furthered the analysis of pigment biochemistry (for reviews see PHILLIPS and FORREST 1980 ; SUMMERS et al. 1982 ; KAYSER 1985 ).

The dull red color of a wild-type Drosophila eye results from the combination of two families of pigment molecules: the pteridines, which are bright red in color, and the ommochromes, which are brown in color. These pigments are deposited in membrane-bounded, protein-containing pigment granules within the pigment cells of the eye and, to a lesser extent, the photoreceptor cells (SHOUP 1966 ; STARK and SAPP 1988 ). Pteridines are synthesized from the precursor GTP, and ommochromes are synthesized from tryptophan. However, our knowledge of pigment synthesis and deposition is incomplete. We do not know the complete reaction sequences of the pigment pathways or the chemical components of the pigment granules. A number of mutations affecting purine metabolism cause a reduction in the synthesis of pteridines (for examples and discussions see NASH and HENDERSON 1982 ; JOHNSTONE et al. 1985 ; HENIKOFF et al. 1986 ). However, these mutations are typically not cell lethal. If cells can externally acquire the nucleotides needed for viability, then it is not clear why they cannot acquire the nucleotides needed for normal pigment synthesis. Finally, many mutations that primarily affect one pigment can also alter levels of pigment in the other family, even though they appear to be synthesized by independent pathways. Thus, the interrelations of pigment biosynthesis remain somewhat mysterious.

The eye of Drosophila has also proven to be an excellent system for developmental biology, especially for studies of cell differentiation and cell-cell communications. An eye consists of ~800 identical repeated structures called ommatidia, which can amplify developmental defects in an ommatidium several hundredfold. Most studies of eye development have focused on the construction of the ommatidia, which make up the majority of the eye. Much less attention has been devoted to uncovering the unique developmental features defining the periphery of the eye (for examples see WOLFF and READY 1991 ; HAY et al. 1994 ). One would expect that a gene differentially affecting the periphery and the middle of the eye would assume a ring pattern of expression. A limited number of mutations that result in a ring pattern of eye pigmentation have been identified. white^halo is an allele of white that produces an eye with a normally pigmented peripheral ring surrounding a lightly pigmented center (JUDD 1975 ; PETERSON et al. 1994 ). The mutations burgundy (a purine auxotroph), doughnut, lozenge^spectacle, and some white transgene insertions show ring patterns or partial rings (arcs) of pigmentation (D. NASH, personal communication; PATTERSON and MULLER 1930 ; WRIGHT 1946 ; JOHNSTONE et al. 1985 ; RUBIN et al. 1985 ; DELATTRE et al. 1995 ; GUBB et al. 1997 ). The cause of the ring patterns has not been identified in any of these cases.

We have isolated a dominant mutation, pugilist^Dominant (pug^D), that differentially affects pigmentation between the margin and the middle of the eye. The pug^D mutation reduces ommochromes and virtually eliminates pteridines in the middle of the eye, while preserving normal pigmentation at the eye margin. Thus, when ommochrome synthesis is blocked by the vermilion or cinnabar mutations, a pug^D/+ eye shows a striking ring of red pigment around the eye margin. The name pugilist was inspired by the similarity to a boxer's black eye. There is also a variegated aspect of the phenotype: the reduction in ommochromes is highly variable, and a small and variable number of cells in the center of the eye show pteridine pigmentation.

The mutation that causes pug^D lies within a gene that encodes the trifunctional enzyme, NADP-dependent methylenetetrahydrofolate dehydrogenase-methenyltetrahydrofolate cyclohydrolase-formyltetrahydrofolate synthetase. This enzyme is referred to as MTHFD, or C1-THF synthase. We use the former designation throughout this article. MTHFD catalyzes interconversion of three derivatives of tetrahydrofolate to provide cofactors for de novo purine biosynthesis (for reviews see BENKOVIC 1980 ; JONES and FINK 1982 ; HENIKOFF 1987 ; APPLING 1991 ). This enzyme is encoded by ADE3 in the yeast Saccharomyces cerevisiae, and mutations in that gene give rise to purine auxotrophy (JONES 1977 ).

In pug^D, a 1-kb piece of highly repetitive DNA is fused to a portion of the coding region of MTHFD. The repetitive segment consists of ~140 iterations of the short sequence AGAGAGA. This sequence is found in high copy number in centric heterochromatin and Y chromosome heterochromatin in D. melanogaster (BONACCORSI and LOHE 1991 ; LOHE et al. 1993 ; for reviews see GATTI and PIMPINELLI 1992 ; LOHE and HILLIKER 1995 ). These repeats are at least partly responsible for the dominant pug^D phenotype. They may also be the cause of the variegated aspect of the pug^D phenotype. The repetitive DNA may be responsible for position-effect variegation (SPOFFORD 1976 ) that occasionally inactivates the dominant mutation.

	MATERIALS AND METHODS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Fly stocks:
Mutations and chromosomes not described here are described by LINDSLEY and ZIMM 1992 . All flies were raised at 25° on standard cornmeal medium. Df(3R)cu Sb/TM6 flies (Df(3R)cu = Df(3R)86C1-2;86D8) were obtained from the Drosophila stock center at Bloomington, Indiana. Df(3R)mbc-R1/TM3, Sb flies (Df(3R)mbc-R1 = Df(3R)95A5-7;95D6-11) were kindly provided by E. Rushton, University of Utah (RUSHTON et al. 1995 ). Flies with the genotype se Ly dn/LVM were provided by the Drosophila stock center at Bowling Green, Ohio.

The original inversion of pug^D is associated with the homozygous lethal mutation, Stubble (Sb). To make homozygous pug^D flies, we screened for double recombinants within the inversion that crossed Sb off the inversion. Virgin females with the genotype v/Basc; +/S²Cyo; pug^D Sb/+ were mated to v; ry males, with 15–20 of both males and females in a bottle. They were transferred to new bottles at 5-day intervals. Out of ~30,000 F₁ progeny screened, 2 males with the genotype v; +/S²Cyo; pug^D/ry were obtained. pug^D homozygous stocks were made from these 2 males. Only one stock was used for future experiments.

Screening for pug^D revertants:
X-ray mutagenesis was carried out in a Torrex 120D X-ray machine. N-ethyl-N-nitrosourea (ENU) mutagenesis was performed as described for EMS by GRIGLIATTI 1986 , except that the neutralization solution for ENU was 1 M NaOH. Males of the genotype v; pug^D cu kar² Sb/TM3, or v; pug^D cu kar² were aged for 3–5 days. They were then either irradiated with 4000 rads or fed a sugar water/ENU solution overnight. The mutagenized males were crossed to v; ry or C(1)DX, y f/Y female virgins, with 15–20 males and 30–40 females in a bottle. After 5 days, the males were discarded and the females transferred to new bottles for further collection. F₁ progenies were identified as potential revertants if they carried the pug^D chromosome 3 and showed uniform eye pigmentation.

Cytology of polytene chromosomes:
Salivary gland polytene chromosomes were prepared as described by LEFEVRE 1976 . For determining the breakpoints of the pug^D inversion, pug^D/+ larvae were used. For cytology of pug^D revertants, chromosomes were from revertant/+ larvae. For in situ hybridization, chromosomes from larvae homozygous for pug^D were used. The chromosomes were prepared as described by PARDUE 1986 . Hybridization and detection were performed using the GENIUS system from Boehringer Mannheim (Indianapolis). Chromosomes were examined with brightfield and phase-contrast optics.

Southern blot analyses and colony hybridization screens:
Fly DNA was purified as described by GOLIC and LINDQUIST 1989 . Southern analyses and colony screens were performed as described by MANIATIS et al. 1982 . Nylon membranes were from Boehringer Mannheim. Hybridization probes were made and hybridization was detected using the GENIUS system and CSPD (Boehringer Mannheim) as a chemiluminescence substrate. A subclone that spanned one of the two inversion breakpoints was identified by the fact that the sizes of the restriction fragments to which it hybridized were different in pug⁺ and pug^D DNA (data not shown).

Cloning and sequencing:
Standard plasmid DNA manipulations were performed as described by MANIATIS et al. 1982 . DNA of P1 genomic clones was prepared by using the QIAGEN Plasmid Midi Kit (QIAGEN Inc., Chatsworth, CA) according to the supplied protocol. P1 clones DS01137 and DS02445 were subcloned into the cloning vector pBluescript II KS (+) (Bluescript) from Stratagene (La Jolla, CA). For the sequence of primers used in PCR analyses and sequencing, see Table 1. For the approximate locations of these primers on the restriction map, see Figure 1.

View larger version (13K): In this window In a new window Download PPT slide   Figure 1. Molecular maps of relevant portions of P1 clones DS01137 and DS02445 from wild-type D. melanogaster. Boxes represent coding regions with direction of transcription indicated by arrows. Vertical arrows mark the inversion breakpoints. Small arrows show the direction (5' to 3') and approximate location of the primers in Table 1, with a number assigned to each primer. Restriction enzymes: B, BamHI; H, HindIII; K, KpnI; R, EcoRI; S, SalI; X, XhoI.

The 5-kb HindIII fragment containing the pug^D breakpoint was cloned directly from genomic DNA. About 200 mg of pug^D genomic DNA was digested with EcoRI, XhoI, and HindIII restriction enzymes. The EcoRI and XhoI digests were to reduce the heterogeneity of the 5-kb HindIII fragments because the 5-kb pug^D junction contains no EcoRI or XhoI sites (results from Southern blot analyses). Digested DNA was phenol-chloroform extracted and run on an agarose gel. DNA of ~5 kb in size was cut out of the gel and purified by using the Geneclean Kit (Bio101 Inc., Vista, CA). DNA was then cloned into the HindIII site of Bluescript. Colonies were screened by hybridization for clones that contained the DNA fragment spanning the proximal breakpoint of the pug^D inversion. The identified clone was labeled p2-1.

DNA clones were sequenced by the Sequencing Facility at the University of Utah. Primers that flank the polylinker sites of Bluescript were used for most sequencing and were provided by the Sequencing Facility (T3, T7, M13 reverse, and M13 forward). Additional sequencing was done using primers homologous to portions of the DNA clones. These were synthesized by the Nucleotide Synthesis Core Facility at the University of Utah. These additional primers were Dist3', DistX5', ProxH2, ProxH3, ProxH5', ProxB3, ProxK1, Rab73', 3'pug-Bst, and MTH3'.

The BLAST program was used to search for homologous sequences for the MTHFD and rab7 genes. Sequences described in this article have been deposited into GenBank under accession numbers AF079459, AF080444, AF080445, AF082097, and AF082098.

Plasmid construction:
DNA fragments were purified from agarose gels (when necessary) using the Geneclean system from Bio101.

The 1-kb AGAGAGA repeats from pug^D are unstable in regular bacterial cloning strains. Bacterial strains we have tried are TOP10 from Invitrogen (Carlsbad, CA), SURE from Stratagene, JM109 from Promega (Madison, WI), and DH5. Plasmids that carried these repeats spontaneously generated DNA clones with shorter repeats during culture growth. We found that growing cells at 30° instead of 37° tended to stabilize the repeats. All pug^D clones were based on the original genomic p2-1 clone that carries the pug^D inversion junction (described above) and on subclones obtained from the P1 clones that cover this region.

Construction of the 14-kb pug^D transgene: The plasmid p2-1 was cut with BamHI and religated. This generated p10-25-15 with a 3.1-kb HindIII-BamHI insert. By a series of cloning steps, we added a 3.9-kb SalI-HindIII fragment, which was derived from P1 DS01137, to the left of the 3.1-kb HindIII-BamHI clone. We also added a 7-kb BamHI fragment, which was derived from P1 DS02445, to the right of this HindIII-BamHI fragment. The correct orientation of the BamHI insert was verified by PCR, using primers Rab75' and Rab73'. The plasmid p11-27-18 gave rise to the expected 700-bp PCR product. p11-27-18 has the 14-kb insert including all three of the potential components of pug^D (Figure 7). This 14-kb SalI-BamHI insert was cloned as a SalI-NotI fragment into the transformation vector pYC1.8, which generated pP[v⁺, pug^D].

View larger version (46K): In this window In a new window Download PPT slide   Figure 2. Phenotypes of pugD. The appearance of the eyes of pug+ and pugD flies in different genetic backgrounds is indicated. The pug genotype is indicated at the top of each column. Eye color mutations carried by the flies are indicated at the left of each row. The flies indicated as whs carry the w1118 null mutation and one copy of a hypomorphic white transgene. Anterior is to the right.

View larger version (69K): In this window In a new window Download PPT slide   Figure 3. The peripheral ring of pteridine pigmentation. Eyes were taken from v; pugD/+ flies. Occasional red spots in the middle are also visible.

View larger version (10K): In this window In a new window Download PPT slide   Figure 4. Cytology of X-ray-induced pugD revertants (rv). The top horizontal line represents the region of the proximal inversion junction in pugD, with numbered and lettered divisions of the chromosome shown. The inversion junction is indicated by a thick vertical line and an arrowhead. The thin horizontal lines represent the portion of the chromosome that remains in each revertant. The extent of each deficiency is marked by two thin vertical lines.

View larger version (132K): In this window In a new window Download PPT slide   Figure 5. Polytene chromosome in situ hybridization with the P1 clone DS01137. Chromosomes were taken from pugD homozygous larvae. DS01137 covers 86C3-6 on a wild-type chromosome 3R. It generates two hybridization signals (arrowheads) on a pugD chromosome. The other inversion breakpoint lies in 95D1-6 region. Cytogenetic map coordinates close to the signals are indicated by arrows.

View larger version (13K): In this window In a new window Download PPT slide   Figure 6. Genomic structure of the pugD inversion. (A) The thick line represents the normal chromosome 3R. The black circle represents the centromere. Close-up views are provided for 86C and 95D regions. Thin horizontal lines in the close-ups represent DNA from the regions. Boxes represent putative transcription units, with directions of transcription indicated by arrows. Vertical arrowheads mark the inversion breakpoints. (B) The proximal inversion junction in pugD with centromere to the left. Hatched box represents the GAGA repeats.

View larger version (15K): In this window In a new window Download PPT slide   Figure 7. Results of transformation experiments. At the top, boxes represent coding regions with direction of transcription indicated by arrows. The shaded boxes below represent the DNA fragments used for transformations. The phenotypes of transformants are indicated at the right. Restriction enzymes: B, BamHI; C, ClaI; H, HindIII; K, KpnI; S, SalI; X, XhoI.

Construction of the pug^D transgene with part of rab7 deleted (a BamHI deletion): The plasmid p11-7-7 lacks the 7-kb BamHI fragment containing the N-terminal two-thirds of rab7 (Figure 7). The 7-kb SalI-BamHI insert from p11-7-7 was cloned as a SalI-NotI fragment into pYC1.8 to produce pP[v⁺, pug(Bam )].

Construction of the pug^D transgene with a KpnI deletion: p11-27-18 with the 14-kb transgene was cut with KpnI and religated. This generated p11-27-18(Kpn ) with a 10-kb insert (Figure 7). The 10-kb KpnI-BamHI insert was cloned into transformation vector pw8, giving rise to pP[w8, pug(Kpn )].

Construction of the transgene with only the GAGA repeats: The original clone p2-1, which has the 5-kb HindIII pug^D junction, was cut with BamHI and religated. This deleted 2 kb of DNA and produced p1-6-7 (same as p10-25-15). p1-6-7 was cut with ClaI and religated. This deleted the start and promoter of MTHFD and left only ~200 bp of MTHFD sequences in the plasmid p1-9-9. The 2.2-kb KpnI-BamHI insert from p1-9-9 was cloned into pw8 to produce pP[w8, GAGA].

Construction of the 3.4-kb KpnI-BamHI pug^D transgene: A 3.4-kb EcoRI-BamHI fragment was isolated from the plasmid pP[w8, pug(Kpn)]. This piece is cloned into pHSS6 (SEIFERT et al. 1986 ). This generates pHSSpug^D. A NotI fragment carrying the transgene from pHSSpug^D was cloned into the unique NotI site in pP[X97]. This gave rise to the transformation construct pP[X97, pug^D]. This construct allows FLP-mediated gene targeting of the pug^D gene (GOLIC et al. 1997 ; see DISCUSSION).

Construction of the wild-type MTHFD transgene: p7-21-18 is an EcoRI subclone of P1 DS01137. A 9-kb KpnI-BamHI fragment from p7-21-18 contains the entire MTHFD⁺ gene. This KpnI-BamHI insert was cloned into pw8 to produce pP[w8, MTHFD].

Construction of the transgene with stop codons upstream of the GAGA repeats: The construct pP[X97, pug^D] has a unique BstEII site 5' of the GAGA repeats with a recognition sequence of 5'-GGTGACC-3'. The construct was cut with BstEII to completion and the ends were filled by the Klenow fragment of DNA polymerase I from Boehringer Mannheim. The treated plasmid was religated. This gave rise to the plasmid pP[X97, pug(Bst^-)] in which the sequence 5'-GGTGACGTGACC-3' was generated at the former BstEII site, with TGA stop codons in two of three reading frames. The sequence changes in pP[X97, pug(Bst^-)] were verified by sequencing.

Construction of a transgene that allows translation of the GAGA repeats and minimal MTHFD sequences: PCR with the primer pair ProxK1 and ATGup generated a 1-kb fragment from pug^D. In ATGup, a BstEII site was introduced just downstream of the ATG codon. The 1-kb fragment was cut with KpnI and BstEII and cloned into pP[X97, pug^D], replacing the 1.2-kb KpnI-BstEII genomic fragment. This generated the construct pP[X97, pug(MTH^-)]. In this construct codons 2–134 of pug^D were deleted. The remaining gene consisted of 44 codons derived from MTHFD along with the GAGA repeats, and could be expressed from the MTHFD promoter. The sequence changes were verified by sequencing.

Construction of a pug^D transgene under the control of the heat shock protein 70 promoter (hsp70): The Drosophila MTHFD protein fails to show extensive amino acid homology to MTHFD from other organisms for the 10–20 amino acids (aa) at the very N terminus. Therefore, the translational start site was predicted as followed. The sequence of the 3.4-kb KpnI-BamHI fragment was examined by computer for promoter and mRNA splice site predictions. These programs are available at the web site of the Berkeley Drosophila Genome Project (http://fruitfly.berkeley.edu/). The first ATG codon lies 125 bp downstream of the predicted transcription start site. We predicted that this ATG codon is the translational start for MTHFD. Additional supporting evidences are as follows: (1) The sequence TATCAAGATG matches the Drosophila initiator ATG consensus: TAAC/AAAA/CATG (CAVENER 1987 ) at 8 of 10 positions. (2) This ATG is in-frame with the codons that show extensive homology to other MTHFD proteins. This was verified by sequencing a putative cDNA clone. We constructed a putative cDNA clone of pug^D based on these predictions.

The pug^D cDNA clone was made by splicing together cDNA sequences and genomic sequences. The 380-bp cDNA fragment from the conceptual translational start site to the HpaI site 95 bp upstream of the GAGA repeats was cloned from a cDNA library. The rest of pug^D cDNA sequences (downstream of the HpaI site) were derived from the genomic sequences of pug^D. Since no introns were predicted between the HpaI site and the stop codons of pug^D, we predict that our clone would have the correct cDNA sequence. The pNB40 Drosophila embryonic cDNA library from N. Brown (University of Cambridge) was used. DNA from the whole library was used as PCR template, with primers 5'pug-K and 3'pug-Bst. In 5'pug-K, a KpnI site was introduced 5' of the ATG. The PCR generated a 530-bp fragment. It was cut with KpnI and HpaI. We replaced the KpnI-HpaI fragment in the pug^D genomic clone with this 380-bp fragment. The 2.8-kb KpnI-BamHI fragment, which contains the pug^D cDNA and its 3'-untranslated region, was cloned into pw8H from K. Basler (BASLER and HAFEN 1989 ). This resulted in pP[w8, hspug^D] that contains a hsp70-driven pug^D gene. The sequence in the putative pug^D cDNA was verified by sequencing.

Drosophila transformation:
All DNA constructs were introduced into the genome by standard P-element transformation (RUBIN and SPRADLING 1982 ). The P-element vectors used for the experiments described here were pw8 (w⁺) (KLEMENZ et al. 1987 ), pYC1.8 (v⁺) (FRIDELL and SEARLES 1991 ), pP[X97] (v⁺) (GOLIC et al. 1997 ), and pw8H (w⁺). The marker genes for the P elements are shown in parentheses. Insertions were mapped by segregation from dominantly marked chromosomes.

Characterization of the ENU-induced pug^D revertant (pug^Drv18):
To detect small deletion(s) possibly associated with the revertant, PCR analyses were performed. DNA was purified from pug^Drv18/Df(3R)cu flies, pug^D homozygotes, and wild-type flies to test DNA of the 86C region. The DNA was used as templates in PCR analyses with primers Test4 and ProxK2. Fragments of 1.9 kb were amplified from all three DNA samples. PCR with primers Test4 and ProxH5' amplified a 0.9-kb DNA fragment from all three DNA templates. DNA was purified from pug^Drv18/Df(3R)mbc-R1 flies, pug^D homozygotes, and wild-type flies to test DNA of the 95D region. The DNA was used as templates in PCR with primers Test1 and LongPCR. All three templates gave rise to a 2.7-kb fragment. PCR with primers Test1 and Rab75' amplified a 1-kb fragment from all three DNA templates.

To demonstrate that the wild-type MTHFD transgene (pug⁺) can rescue the mutant phenotype of pug^Drv18/Df(3R)cu flies, the following crosses were performed. Female virgin flies of w¹¹¹⁸ P[w8, MTHFD]11A, which carry a pug⁺ transgene within a P element inserted on X, were mated to males of pug^Drv18/Df(3R)cu. As Df(3R)cu/+ flies have a dominant Minute phenotype (short and thin bristles), Minute⁺ male progeny have the genotype of w¹¹¹⁸ P[w8, MTHFD]11A; pug^Drv18/+. They were mated to virgin females of pug^Drv18/Df(3R)cu. As Df(3R)cu carries the recessive cu mutation, and the original pug^Drv18 chromosome also carries cu and a linked recessive lethal mutation, pug^Drv18/Df(3R)cu flies have curled wings and pug^Drv18/pug^Drv18 flies are dead. Male progeny that were pug^Drv18/Df(3R)cu did not carry the pug⁺ transgene on X. They all showed the recessive eye pigment phenotype. Female progeny that were pug^Drv18/Df(3R)cu were heterozygous for the pug⁺ insertion on X.

To generate flies homozygous for pug^Drv18, females that were pug^D/pug^Drv18 were generated. Homozygous stocks were made from recombinants that had crossed off the lethal mutation on the pug^Drv18 chromosome.

Heat shock experiments:
Because animals with a single copy of the construct pP[w8, hspug^D] show subtle pigment defects after heat shock, we used animals with multiple copies of the gene to increase expression of pug^D after heat shock. The hspug^D transformants are marked by a hypomorphic white gene (w^hs). Flies with one copy of w^hs usually have orange or yellow eye color, whereas flies with two copies have red eyes. On the basis of this phenotype, recombinants were made so that one chromosome 3 harbored two hspug^D transgenes. The hspug^D insertions that were used in the heat shock experiments are as follows, with chromosomal locations in parentheses: 7A (X), 8A (III), and 9A (III). The flies with the genotype w⁺ P[w8, hspug^D]7A were made by crossing virgin females of w¹¹¹⁸ P[w8, hspug^D]7A/+ to wild-type males. The progeny were heat shocked at early to mid-pupal stage. Male recombinants that show pigment defects (see RESULTS) should have the genotype of w⁺P[w8, hspug^D]. Two such males were recovered. Flies that are w⁺ P[w8, hspug^D]7A; P[w8, hspug^D]8A P[w8,hspug^D]9A/TM3 were heat shocked to generate the eyes shown in Figure 9.

View larger version (60K): In this window In a new window Download PPT slide   Figure 8. Phenotypes of pugD transformants. This figure shows a sample of the phenotypic variation observed with independent insertions of the 14-kb S-B transgene (Figure 7). All flies were homozygous for the recessive cn mutation and heterozygous for a pugD insertion.

View larger version (82K): In this window In a new window Download PPT slide   Figure 9. Phenotype of flies carrying the construct pP[w8, hspugD]. Eyes are from otherwise wild-type males that carried three copies of the hspugD transgene (for the genotype see MATERIALS AND METHODS). Flies were heat shocked during early to middle pupal stage.

To determine the timing of pug^D action, flies were allowed to lay eggs until pupae with black wings were present in the vials. Parents were then removed and the vials were heat shocked in a water bath at 38° for 1 hr. The adults that eclosed were examined for pigment defects on the day that they eclosed.

Images:
Images were obtained and processed as described (GOLIC 1994 ; AHMAD and GOLIC 1996 ). For Figure 2, Figure 3, Figure 8, and Figure 9, the heads from 3-day-old flies were used.

	RESULTS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Discovery of pug^D and basic phenotypes:
The dominant mutation pug^D was discovered in a screen for X-ray-induced chromosomal rearrangements that caused position-effect variegation of a white transgene (w^hs) located on chromosome 2 (AHMAD and GOLIC 1996 ). In one case, when a fly with variegated eye pigmentation was outcrossed to flies with the w¹¹¹⁸ null mutation, some of its progeny showed uniformly pigmented eyes. In other words, the genetic element responsible for causing variegation segregated from the w^hs transgene. The mutation that caused the variegation was mapped to chromosome 3.

We wished to determine whether this mutation produced variegation by a specific interaction with the w^hs transgene, or whether variegation also occurred in w⁺ flies. Flies with the mutation were crossed to flies with different insertions of the same transgene and to w⁺ flies. Variegation of eye pigment was observed in all cases (Figure 2). Because the w^hs transgene and the w⁺ gene have different 5' regulatory sequences (KLEMENZ et al. 1987 ), it is unlikely that this mutation is exerting its effect through transcriptional regulation of the white gene.

Many mutations that affect eye pigmentation have defects primarily in one of the two major pigment pathways. These pathways can be knocked out independently by mutations in the vermilion (v) and the brown (bw) genes. We placed the pug^D mutation (as pug^D/+) in separate backgrounds of v and bw to determine whether it affected one or both pigments. Homozygous bw flies produce only ommochrome pigment, and pug^D causes a variegated reduction in ommochrome pigmentation (Figure 2). A much more dramatic effect is seen in flies that can only synthesize pteridines because they carry the v mutation. In a v background, pug^D almost completely eliminates pteridines in the middle of the eye (Figure 2). However, the periphery of the eye remains almost completely pigmented. The result is a striking ring of pigment surrounding a center that is almost completely white. When we tilt the eye to look at the eye margin at an angle that is perpendicular to the eye surface, it can be seen that the peripheral pigmentation is confined to the very edge of the eye (except for the occasional spots) and appears to be external to all ommatidia (Figure 3). There is also a variegated aspect of the pteridine pigmentation in pug^D: small and infrequent red spots can be seen in the middle of a v; pug^D/+ eye (Figure 2 and Figure 3). A small portion of the pug^D flies have concave dents in the eye as if the overall eye structure is weak (not shown).

Although the pug^D mutation is dominant, homozygous pug^D flies survive (see MATERIALS AND METHODS). The phenotype of homozygotes is almost identical to that of heterozygotes, with only a slight lessening of pigmentation. It is likely that this reduction in pigment is attributable to homozygosity for the recessive karmoisin² mutation that the pug^D chromosome carries. Some pug^D homozygotes also have slightly rough eyes (not shown).

Mapping the pug^D mutation:
The pug^D mutation is associated with an inversion on the right arm of chromosome 3 (3R), with breakpoints at polytene chromosome bands 86C3-4 and 95D1-6. A simple inversion involves two breakpoints and creates two new DNA junctions. Therefore pug^D could be located at either junction. It is also possible that pug^D is unrelated to the inversion, and their occurrence together may be merely coincidental.

To map pug^D, we screened for X-ray-induced revertants of pug^D. Because deficiencies of either the 86C or 95D regions, carried in Df(3R)cu/+ and Df(3R)mbc-R1/+ flies, respectively, do not produce eye pigment phenotypes, pug^D is not likely to be the result of haplo-insufficiency. Therefore we expected that the dominant phenotype could be reverted by deleting the pug^D gene.

Seventeen phenotypic revertants were recovered among ~70,000 F₁ progeny of irradiated males (half of which received the pug^D chromosome). Lines were established from 10 of the revertants, and their polytene chromosomes were studied to determine the nature of the reversion. In 9 out of 10 cases, we found a small deletion of the proximal inversion junction. The tenth revertant was associated with a T(Y;3), with the chromosome 3 breakpoint at the proximal junction of the inversion. Figure 4 summarizes the cytology of the revertants. We conclude that pug^D is at or close to the proximal inversion junction.

We also recovered one male in which pug^D variegates as a result of insertional translocation of the pug^D region into the heterochromatic Y chromosome. The eyes of v; +/Tp(3;Y), pug^D males show red sectors in the middle of the eye on an otherwise white background. The peripheral ring of pigment is visible in the white portions of the eye (not shown). This appears to be a typical case of position-effect variegation in which the euchromatic pug^D gene has been placed close to heterochromatin and experiences variegated inactivation, producing the red sectors.

Because the aneuploid segregants of this transposition survive, we were able to test the effect of the pug^D mutation in flies with two copies of pug⁺ gene. In these v/Dp(3;Y), pug^D ; pug⁺/pug⁺ flies, the pug^D phenotype is still visible and essentially identical to the phenotype of v; +/Tp(3;Y), pug^D flies. Therefore, the pug^D phenotype cannot be suppressed by a second copy of pug⁺, providing further evidence that the phenotype does not result from a deficiency of the wild-type product.

Cloning and sequencing of the pug^D junction:
Wild-type P1 genomic clones derived from the 86C and 95D regions of chromosome 3 were obtained from the Drosophila Genome Center. Polytene chromosomes of pug^D homozygotes were hybridized in situ with labeled probes made from the P1 clones. If a P1 clone spans one of the two breakpoints of the inversion, it should generate two hybridization signals on chromosome 3R from pug^D homozygotes. We found two clones that fulfill this condition (one example is shown in Figure 5; for restriction maps see Figure 1).

These two P1s were further subcloned and Southern blots were used to look for restriction fragment length polymorphism by hybridizing digested DNA from pug⁺ and pug^D flies with probes from either the whole P1 clone or individual subclones. The breakpoints were thus mapped to individual subclones. The pug^D junction is located within a 5-kb HindIII fragment. The DNA surrounding this junction was sequenced to determine the genetic architecture of the region.

Figure 6 and Figure 7 depict the genomic structure of the pug^D junction. At chromosome region 86C, the inversion breakpoint lies within an open reading frame of a gene that is highly homologous (~60% amino acid identity) to genes from a number of organisms that encode the trifunctional MTHFD enzyme. This MTHFD-homologous gene is different from that found by PRICE and LAUGHON 1993 at 85C. The gene they reported appeared to encode the mitochondrial version of the enzyme. The gene at 86C appears to encode the cytoplasmic form. In chromosome region 95D, an open reading frame ~400 bp away from the breakpoint encodes a protein with ~90% amino acid identity to the rab7 gene from a variety of other organisms. SATOH et al. 1997 have identified a fragment in Drosophila with ~91% amino acid identity to rab7 genes from other organisms. As their sequences were not published, we could not compare the rab7 we identified with their fragment.

The sequence of the distal breakpoint (non-pug) showed a perfect rejoining of the inversion breaks. However, the HindIII fragment that contains the proximal inversion breakpoint is 1 kb larger than the 4-kb HindIII fragment predicted by a simple breakage-and-fusion event. To determine the nature of the extra DNA, we cloned the 5-kb piece directly from genomic DNA of pug^D flies (see MATERIALS AND METHODS). A 1-kb piece of highly repetitive DNA has been inserted at the proximal inversion junction, accounting for the increased size of the HindIII fragment. This DNA is not normally present in the vicinity of the MTHFD or rab7-homologous genes. The repetitive DNA consists of ~140 units of AGAGAGA repeat (GAGA repeats) with occasional slight variations. The same repeats have been identified as a major component of Drosophila heterochromatin DNA (for references see the Introduction).

In summary, the pug^D junction consists of the N-terminal one-fifth of a gene that appears to encode MTHFD, 1 kb of AGAGAGA repeats fused to this gene, and a rab7-homologous gene 400 bp distal to the junction (Figure 7).

Transformation of the pug^D gene:
To identify pug^D, DNA from the proximal inversion junction was transformed into pug⁺ flies. A 14-kb DNA clone was constructed by splicing together wild-type genomic subclones of the region and the 5-kb HindIII fragment that contains the pug^D junction (Figure 7). This was placed in a P-element vector for germline transformation. Fifteen independent transformants were isolated. All transformants showed pigment defects in a wild-type background (not shown), similar to the phenotype shown in the top right of Figure 2. Because the P-element vector carried a v⁺ gene as a transformation marker, it was not possible to examine the phenotypes of transformants in a v background. However, the cinnabar (cn) mutation also eliminates ommochrome pigments, and a cn; pug^D/+ fly has the same pattern of pigmentation as a v; pug^D/+ fly (not shown). We crossed the transformants into a cn background and found that the original pug^D phenotype was reproduced (Figure 8). Thus, pug^D is contained within this 14-kb segment. The transformed pug^D gene also caused a variegated reduction in ommochrome pigmentation, similar to the original pug^D mutation (not shown).

Because two genes are present in this fragment of DNA, we subcloned and transformed flies with portions of the 14-kb DNA fragment to pinpoint the responsible gene. A smaller, 7-kb SalI-BamHI fragment that contains only the C-terminal one third of rab7 is still capable of generating the pug^D phenotype (Figure 7). We conclude that rab7 is not the cause of pug^D. Furthermore, DNA between the SalI and KpnI sites is not necessary for the phenotype, because a 10-kb KpnI-BamHI fragment can also reproduce pug^D (Figure 7). The pug^D gene is entirely contained within the smaller 3.4-kb KpnI-BamHI fragment. This was verified by transformation (Figure 7). Finally, a construct that lacked most of the remaining MTHFD sequences, including the start codon and upstream sequences, was also transformed (the 2.2-kb ClaI-BamHI fragment in Figure 7). None of the 13 transformants showed the pug^D phenotype. Therefore, pug^D is a mutation in a gene that appears to encode the enzyme MTHFD, and it consists of the DNA that codes for the MTHFD N terminus fused to AGAGAGA repeats.

Null mutations of pug have a recessive eye color phenotype:
In other experiments, a pug^D revertant allele (pug^Drv18) was obtained by treatment with the chemical mutagen ENU. Cytological analyses of polytene chromosomes (not shown), PCR (see MATERIALS AND METHODS), and Southern blot analyses (not shown) revealed that it has no detectable deletion at pug^D. When the revertant allele was heterozygous with Df(3R)cu, which deletes the whole MTHFD-containing region of 86C, we observed a recessive phenotype slightly reminiscent of pug^D. In pug-null flies that are otherwise wild type, the center of the eye is dully pigmented, with a bright ring toward the periphery. This eye color phenotype shows up only in young flies. We believe that the center of the eye is less pigmented than the periphery. Because it has less pigment, it is less effective at reflecting light to the observer and therefore appears darker.¹ This recessive phenotype was rescued by a P-element insertion carrying pug⁺ (the wild-type MTHFD), thus identifying a defect in pug as the cause of this recessive phenotype. We generated pug^Drv18 homozygotes (see MATERIALS AND METHODS), and these flies show the same recessive eye color phenotype as pug^Drv18/Df(3R)cu flies. This suggests that pug^Drv18 is likely to be a null allele of MTHFD.

We also studied the effect of pug-null on ommochrome and pteridine levels separately. A bw; pug^Drv18/Df(3R)cu eye does not show a reduction in pigmentation when visually compared with a bw eye (not shown). This suggests that pug-null does not affect ommochrome pigmentation. A very weak eye color defect can be seen in a v; pug^Drv18/Df(3R)cu eye. Eyes from some of the very young flies show a lighter eye center than the periphery. Therefore we believe that the pug-null mutation slightly reduces pteridines in the eye center. Moreover, this eye color defect is very similar to phenotypes of other mutations that affect purine de novo synthesis: it reduces pteridine pigmentation initially, but the phenotype wanes as flies age (JOHNSTONE et al. 1985 ; KEIZER et al. 1989 ; TIONG and NASH 1990 ).

A mutation that had a phenotype similar to the recessive pug phenotype was previously mapped in the vicinity of the pug gene (WRIGHT 1946 ). The doughnut (dn) mutation was located at map position 50 on chromosome 3. This region corresponds approximately to the polytene chromosome region 86C, where pug is located. To test for allelism we obtained a dn stock from the Mid-America Drosophila stock center. This stock was reported to carry the dn allele heterozygous with a balancer chromosome. In crosses of the putative dn stock to flies with deletions of pug^D (Figure 4) and to Df(3R)cu/+ flies, none of the dn/Df flies showed an eye color phenotype. This would normally lead to the conclusion that pug and dn are not allelic. However, animals with the putative dn-bearing chromosome did not survive as homozygotes, so it was not possible to verify the presence of the mutant dn allele. Therefore, we cannot with certainty state that pug^D and dn are not alleles of the same gene.

Translation of the GAGA repeats is necessary for the dominant phenotype:
By DNA sequencing, no stop codons were found upstream of the GAGA repeats or within the sequenced portion of the repeats (about 700 bp). Stop codons are present in all three reading frames immediately downstream of the GAGA repeats. Therefore, it is likely that the 1-kb repeats are translated in pug^D. To test this hypothesis, we made a construct, pP[X97, pug(Bst^-)], in which the unique BstEII site was cut and filled. This treatment introduces stop codons in two of the three reading frames. It also caused a +1 frame shift in the MTHFD coding frame. The new frame stops 16 codons downstream without reaching the GAGA repeats. This altered pug^D gene is predicted to code for a short peptide of 153 amino acids (aa), in which 137 of them are from the N terminus of MTHFD, and the extra 16 aa are the result of the +1 frame shift. Six independent transformants of pP[X97, pug(Bst^-)] were recovered: none showed any pigment defects even when homozygous for the transgene.

In the pug^D protein, there are 178 codons upstream of the GAGA repeats. Only the first 137 would be translated in pP[X97, pug(Bst^-)]. One might still argue that synthesis of all 178 is necessary for the phenotype. However, a transgene with shortened repeats failed to reproduce the dominant phenotype. Because the GAGA repeats are somewhat unstable in regular bacterial cloning strains, we recovered a spontaneous derivative of pP[X97, pug^D] with an internal deletion, which we named pP[X97, pug^S]. The deletion shortened the GAGA repeats from 1 kb to ~300 bp, and restriction mapping indicates that this is the only change. The first 178 codons should be translated with this transgene, but it did not produce the pug phenotype. This strongly suggests that translation of these 178 codons is not sufficient for the phenotype—translation of a long stretch of GAGA repeats (>300 bases) is also needed to produce the phenotype. However, we cannot at this time exclude the alternative explanation that the pug^S mRNA is destabilized by the loss of 700 bp of repeats, and this causes reversion of the pug phenotype.

Another possible cause for the failure of both pP[X97, pug(Bst^-)] and pP[X97, pug^S] constructs to regenerate the pug phenotype is that the chromosomal positions of the insertion do not allow sufficient expression of pug^D. To address this question, we used the FLP-mediated DNA mobilization technique (GOLIC et al. 1997 ) to target various transgenes to specific chromosomal sites. At one particular site, a pug^D transgene with the full-length GAGA repeats and no stop codons upstream of the repeats (the 3.4-kb K-B fragment of Figure 7) generated an eye color phenotype essentially identical to the original pug phenotype. However, when we targeted to the same chromosomal site, either the pug^D transgene with a premature stop codon (as in pP[X97, pug(Bst^-)]), or the transgene with shorter repeats (as in pP[X97, pug^S]), the pug phenotype was not produced. These experiments eliminate position effects as the cause for the failure of both pP[X97, pug(Bst^-)] and pP[X97, pug^S] to produce the pug phenotype. Therefore, the DNA sequence changes introduced in vitro knocked out the pug^D gene. The predicted pug^D protein would consist of the N-terminal 178 aa from MTHFD and a C-terminal ~350 aa in which the sequence SLLSSLF is repeated ~50 times with occasional substitutions of C, V, and P (data not shown).

We also wished to determine if translation of the repeats alone would produce the pug phenotype. The construct pP[X97, pug(MTH^-)] should allow translation of the full-length GAGA repeats plus 44 codons of MTHFD (see MATERIALS AND METHODS). None of the three transformants with this construct show any pigment loss.

Developmental timing of the pug^D effect:
To determine when in development the pug^D gene exerts its effect on pigmentation, we placed the predicted pug^D cDNA under the control of the hsp70 promoter (see MATERIALS AND METHODS). We then heat-shocked animals with the hspug^D construct at different developmental stages to induce pug^D expression. Flies with a single hspug^D gene show pigment defects when heat-shocked at early to middle pupal stage, exhibiting occasional small spots that lack pigment (not shown).

To increase the level of hspug^D expression, we generated flies with three copies of this transgene. These flies showed much greater loss of pigment after heat shock (Figure 9). They were used to determine the timing of pug^D action. Animals were given a single 1-hr heat shock at 38°. Flies that eclosed during the third day after heat shock showed pigment loss. Thus, the most sensitive period for pug^D expression is 3 days before eclosion or during the second day of pupal development. The eye pigments are first visible in wild-type flies around 48 hr after puparium formation (PHILLIPS and FORREST 1980 ). Therefore, pug^D acts near the time when eye pigments are first made.

	DISCUSSION

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

We have discovered an eye color mutation in Drosophila melanogaster with two unusual characteristics. First, the pug^D mutation causes a dominant and variegated reduction in ommochromes and pteridines throughout the eye, but null alleles of this gene are not dominant. Before this discovery, there existed only a few examples of such a dominant variegating eye color mutation (for example, bw^D). Second, pteridine pigmentation is seemingly unaffected around the periphery of the eye, leaving a ring of pteridine pigment. Examples of ring patterns of pigmentation are rare, and none are well understood. Thus it seemed that much could be learned by identifying and characterizing the mutation that caused these phenotypes. Genetic and molecular methods have been employed to achieve this goal.

The nature of the pug^D mutation:
The pug phenotype does not result from a simple change in gene dosage. When one copy of pug⁺ is deleted, flies do not exhibit the pug phenotype, and pug^D is not suppressed by three copies of pug⁺ (not shown). Thus, haplo-insufficiency is ruled out as the cause. It is equally unlikely that the pug^D rearrangement causes overexpression of pug⁺ on the homolog (possibly by some sort of transvection effect, e.g., GEYER et al. 1990 ) and that this overexpression produces the pug phenotype. Two extra copies of a pug⁺ transgene do not produce the pug phenotype (not shown), and simple deletions of the pug^D breakpoint that leave pug⁺ on the homolog intact revert the phenotype. Further, pug^D hemizygotes still exhibit the pug phenotype. The pug phenotype is caused by the novel fusion gene present at the pug^D inversion breakpoint. P elements that carry this fusion gene confer the pug phenotype, and, when the coding sequence of this fusion gene is placed under control of the hsp70 promoter, heat shocks can partially reproduce the pug phenotype.

The pug^D mutation lies within a gene that, based on sequence similarity, encodes the enzyme MTHFD. The activities of this enzyme are essential for de novo purine synthesis. The intimate relationship between purine metabolism and pteridine synthesis is apparent from the fact that pteridines are synthesized from GTP. Many mutations that affect de novo purine synthesis also reduce the level of pteridines in the eye (JOHNSTONE et al. 1985 ; TIONG et al. 1989 ; TIONG and NASH 1990 ; CLARK 1994 ). Thus, it is possible that pug^D causes loss of pteridine pigment by interfering with purine biosynthesis.

The pug^D allele is a fusion gene in which the N-terminal one-fifth of the MTHFD coding region (178 codons) is joined, within the coding region, to 1 kb of highly repetitive DNA. The fusion gene was created by an X-ray-induced chromosomal inversion, with the repetitive DNA (presumably originating from centric heterochromatin) captured between the normal euchromatic sequences of the proximal breakpoint. Transcription and translation of the repeats are apparently necessary to produce the dominant phenotype, because insertion of a stop codon in the mRNA just upstream of the repeats abolishes the dominant phenotype. The repetitive DNA consists of repeats of the sequence AGAGAGA, with occasional slight variations. This part of the gene would encode a sequence of the amino acids SLLSSLF, repeated ~50 times with occasional substitutions of C, V, and P. Because of the high leucine content, this region of the protein has the potential to form leucine zipper motifs, and to thereby promote homotypic protein-protein interactions (reviewed in VINSON et al. 1989 ; BUSCH and SASSONE-CORSI 1990 ). Stop codons are present in all three reading frames a short distance downstream of the repetitive DNA. Thus, this repetitive peptide stretch should constitute two-thirds of the mutant protein, at the C terminus.

The trifunctional MTHFD enzyme generates 10-formyltetrahydrofolate as a cofactor for two methyltransferases in purine biosynthesis [for reviews of folate-dependent enzymes see BENKOVIC 1980 and APPLING 1991 ; for reviews of purine synthesis see JONES and FINK 1982 and HENIKOFF 1987 ]. MTHFD contains two functional domains. The dehydrogenase and cyclohydrolase activities reside in the N-terminal one-third, while the synthetase activity is associated with the C-terminal two-thirds (PAUKERT et al. 1977 ; TAN and MACKENZIE 1977 ). The MTHFD protein appears to form homodimers and has been suggested to participate in the formation of a complex of enzymes for purine biosynthesis in yeast, with both domains of the protein involved in interactions with other members of the complex (APPLING 1991 ; WEST et al. 1996 ).

The mechanism of pug^D action:
One hypothesis for pteridine elimination in pug^D is that the mutant pug^D protein disrupts pigment granules. In Drosophila, ommochromes and pteridines are present in membrane-bounded protein-rich pigment granules (SHOUP 1966 ). Many enzymatic activities for pteridine biosynthesis have been shown to be closely associated with granules (DORSETT et al. 1979 ; HEARL and JACOBSON 1984 ). Purified granules are able to produce drosopterin, a major component of the eye pteridines. These results led to the suggestion that pigment granules contain all enzymes for drosopterin synthesis. The mutant pug^D protein might also be located in pigment granules. The C terminus of pug^D is expected to be hydrophobic and it might disrupt pigment granule structure or stability. It has been suggested that the white protein is located across the membrane of pigment granules (TEARLE 1991 ; MONTELL et al. 1992 ). The pug^D protein could disrupt both pteridine and ommochrome pigmentation by interfering with the function of the white protein in pigment granules. The pug^D protein might also bind the pterin ring of the pteridines, as the pterin ring is a part of the tetrahydrofolate molecule [for chemical structures of pteridines and folates see TEMPLE and MONTGOMERY 1984 ; PFLEIDERER 1993 ]. It may then disrupt normal pigment synthesis either by binding and sequestering pigment precursors, or by interfering with other enzymes in the pteridine pathway. This model, although simple, does not take into account the normal function of MTHFD in purine synthesis, nor does it connect the phenotypes of the null and dominant alleles of pug. Therefore, we do not favor this model.

A second model for pug^D action is based more directly on the supposed role of pug⁺ in purine synthesis. The pug^D protein might act as a toxic subunit in a multimeric assembly, crippling purine biosynthesis, and thereby reducing pteridine synthesis. The mutant protein may interact with the wild-type MTHFD protein or other members of a purine synthesis protein complex, and by this route, poison the complex. This model predicts that extra copies of pug⁺ should suppress pug^D, but they do not. If the pug^D peptide were acting as a toxic subunit, pug⁺ subunits should compete for its location in a multimeric assembly. Thus, we do not favor this model. However, the mutant pug^D gene may be expressed at a much higher level than the normal level of the pug⁺ gene so that the mutant protein is in great excess to the wild-type MTHFD, both in pug^D/+ heterozygotes and in pug^D transformants. If true, a manyfold increase in the amount of MTHFD may be necessary to generate an observable suppression of the pug phenotype. Therefore, we cannot completely discard this model. However, we can rule out the hypothesis that pug^D affects pigmentation by an interaction with the wild-type MTHFD protein, because pug^D/pug^D and pug^D/Df flies both exhibit the pug phenotype, but no pug⁺ gene is present.

A third model supposes that the N-terminal remnant of MTHFD synthesized by pug^D may irreversibly bind and sequester its tetrahydrofolate substrate. If this substrate is present in a limiting concentration, it may be reduced to a level that cannot support the pug⁺-mediated synthesis of the cofactor used to make purines and thus prevent pteridine synthesis. This model predicts that the phenotype of pug-null flies with respect to pteridine pigmentation should be the same as that of pug^D, or more severe than pug^D, and this was not observed. Although pug^Drv18/Df flies exhibit a phenotype that is slightly reminiscent of pug^D, it is much less severe than the pug^D phenotype. This objection might be removed if there was another enzyme capable of providing the tetrahydrofolate cofactor for purine synthesis. This is true in the yeast S. cerevisiae. The MTD1 gene in yeast encodes a cytoplasmic NAD-dependent methylenetetrahydrofolate dehydrogenase that can provide 10-formyltetrahydrofolate for purine synthesis (WEST et al. 1993 , WEST et al. 1996 ). This enzyme is monofunctional and shows some amino acid similarities to the trifunctional MTHFD. In Drosophila, there might be an enzyme with similar activity so that some level of cofactors could be maintained to support purine synthesis even in the absence of the MTHFD proteins. Null alleles of genes in purine de novo synthesis behave as lethals or extreme semilethals, with characteristic late pupal death (TIONG et al. 1989 ; TIONG and NASH 1990 ). Genetic and molecular analyses suggest that the pug^Drv18 mutation is likely a null. However, pug^Drv18/Df(3R)cu flies survive very well, indicating functional redundancy in providing the essential folate cofactor for purine synthesis. In pug-null flies it is likely that de novo purine synthesis still occurs, though at a reduced level, and this allows for near-normal pteridine synthesis.

This theory can explain why the dominant defect is more extreme than the recessive defect: the dominant mutation cripples both modes of synthesis for the purine pathway cofactor, while the recessive allele affects only one. However, it still seems that viability should be reduced by the dominant mutation if it causes a severe defect in purine synthesis. This conceptual difficulty could be overcome if the expression of pug^D were limited to the eye. Flies with pug^D do sometimes exhibit phenotypes that could be attributed to an impairment of differentiation or to cell lethality owing to a defect in purine synthesis in the eye. Some pug^D/+ flies seem to have structurally weak eyes, and occasionally homozygotes have rough eyes (not shown). If these defects were extended throughout the body, lethality might result. In flies with the hspug^D transgenes we did observe some disruption of body patterns after heat shocks, possibly as a result of cell death induced by pug^D throughout the body.

The last two models, which propose that the defect in pteridine pigmentation in pug^D stems from a defect in purine synthesis, can also be invoked to account for the ring pigmentation of pug^D.

Patterned pteridine pigmentation in pug^D:
One effect of pug^D is to eliminate pteridine pigment from the center, but not the margin, of the eye. This pattern might also be a simple consequence of a defect in purine synthesis. If the failure to make pteridines in pug^D is caused by purine deficiency, then externally supplied purines should suppress the phenotype. Although Drosophila can utilize dietary purine, the eye pigmentation process occurs in the pupal stage when absorption of exogenous nutrients is blocked by the pupal case. Therefore, an increased demand for purines (for example, to make pigment) would have to be met either by an increased rate of de novo purine synthesis or by purine uptake from neighboring cells or hemolymph. In fact, it has been shown that externally supplied guanine derivatives are actively taken up by the eye, and the transported guanine compounds are converted to pteridines (MONTELL et al. 1992 ). It has also been shown that guanine derivatives can travel between pigment-producing cells in Drosophila (SULLIVAN et al. 1979 ; MONTELL et al. 1992 ). Because of the hemispherical shape of the eye, only those cells at the periphery have substantial surface area in contact with noneye tissue (Figure 10). Perhaps then, only at the eye margin would exogenous purines be available to cells of the eye. In a pug^D fly these would be the only cells able to make pteridine pigment, and this would generate the ring of pteridine pigment around the eye. Moreover, if pug^D expression is eye specific, purine levels could be reduced in the eye but not in the surrounding tissues. This would create an even greater difference in purine concentration that would certainly facilitate purine transport into the eye.

View larger version (37K): In this window In a new window Download PPT slide   Figure 10. A model for the pugD pigment ring. Frontal (A and B) or lateral (C and D) views of the phenotypes of v (A and C) and v; pugD/+ (B and D) are diagrammed. A sectional view of one eye is provided for v and v; pugD/+ (E and F, respectively). The plane of section is indicated with arrows in C and D. In E and F the bundles of retinula cells and their overlying lenses are indicated as cones. The pigment cells lie between adjacent bundles of retinula cells. In a pug+ eye (E) purine uptake occurs at the periphery of the eye (arrows), and purine uptake combined with endogenous synthesis is sufficient to produce pigment in all pigment cells. In a pugD eye (F), endogenous purine synthesis does not occur. Purines are still taken up at the edge of the eye, but are used to produce pigment in only the outer layer of pigment cells and are not transported further toward the center of the eye.

An alternative explanation for the differential pigmentation in pug^D is that the mutant gene product is present throughout the eye, except at the periphery. Regulatory elements that allow spatially controlled expression of a gene are common in Drosophila. For example, tissue-specific enhancers have been found for the white gene (LEVIS et al. 1985 ; PIRROTTA et al. 1985 ) and the yellow gene (GEYER and CORCES 1987 ; MARTIN et al. 1989 ). Further, even within an apparently uniform tissue such as the eye, a variety of patterns of gene expression can be observed (SUN et al. 1995 ). Therefore, it is tempting to imagine that regulatory elements within the pug^D fragment carried by the 3.4-kb transgene (Figure 7) provide for pug^D expression in the middle of the eye but not at the eye margin. These could be part of the normal regulatory elements for either the MTHFD or rab7 genes. It should be straightforward to identify the proposed regulatory elements for pug^D by performing further deletion analysis of the 3.4-kb construct. However, we cannot rule out the possibility that the AGAGAGA repeats themselves could constitute a signal for exclusion of pug^D from the eye margin. Because these repeats must also be translated to produce the pug phenotype it may be difficult to establish any role they have in regulation.

The effect of pug^D on ommochrome synthesis:
Because pug^D has a very strong effect on pteridines, we have focused our discussion on explanations for that defect. However, pug^D belongs to a class of mutations that affect both pigments of the Drosophila eye (PHILLIPS and FORREST 1980 ; FERRE et al. 1986 ). The very existence of this class of mutations points to an inter-relation between ommochrome and pteridine synthesis. Different models to explain this interrelation have been entertained by SUMMERS et al. 1982 . The general belief is that the two pathways share common systems for precursor transport and storage. At the ultrastructural level, it has been suggested that the presence of pteridines in Drosophila eyes is needed for normal ommochrome granule formation (SHOUP 1966 ; FUGE 1967 ; SUMMERS et al. 1982 ; REAUME et al. 1991 ). Therefore the reduction of ommochromes in pug^D may be a s